Serenity, the stock-picker whose reconstructed views this site tracks, spent Friday pushing back against what he framed as an overblown selloff in the AI optical-interconnect space — the corner of the market he cares about most.

"Nothing has changed." His day began with a defense of the photonics complex, arguing that two catalysts behind the recent weakness — a Bloomberg report on Meta's compute plans and separate reports of delays in NVIDIA's roadmap — had already been walked back by internal memos and denials. His point: the fundamentals underpinning his holdings are intact even as prices swing. He ticked through the names. Lumentum (LITE), the large-cap optical incumbent he uses as a quality benchmark, remains sold out on capacity for years. Sivers (SIVE), the tiny Swedish maker of indium-phosphide continuous-wave lasers (the light source for co-packaged optics) that is his single largest conviction bet, is ramping volume with partners like GlobalFoundries, Jabil, Poet and Ayar. Applied Optoelectronics (AAOI), his Made-in-America transceiver play, still carries its aggressive H2-2027 revenue targets. IQE (IQE), the UK epiwafer foundry he holds as a turnaround, keeps its MACOM and Tower Semi contracts. And AXT (AXTI), which he argues controls roughly 40% of the InP substrate supply chain, hasn't lost that grip. TSMC (TSM), which he treats as the indispensable foundry at the center of the whole AI trade, rounded out the list. His broader message: thematic selloffs — like the 2025 space-sector and neocloud drops — rarely reflect individual company health.

A short-seller tell. Serenity flagged a Rosenblatt note on the optical sector's poor two-month run, blamed on CPO-delay reports and China capacity fears. The kicker he seized on: short-sellers reportedly told the analysts they don't strongly believe the bear thesis and plan to close positions in late July/early August, then buy optical names for the 2027 CPO scale-up. He read that as confirmation that institutions simply want a later entry point. He tied a second Rosenblatt note to his core Sivers case — the argument that even if optimizations cut transceivers per chip, surging XPU counts keep total optical demand rising, and that ultra-high-power CW lasers are the ideal CPO architecture (citing Broadcom). That directly supports his light-source thesis and, by extension, Sivers' Nasdaq-tracked line (SIVEF) and Coherent (COHR), the vertically integrated photonics compounder he holds as a lower-beta anchor.

Odds and ends. Serenity noted Sivers rallied, joking that the move (and the CEO's open-market buying) had been "predicted" by an invented technical-analysis indicator — a jab at chart-driven retail conviction. He also tracked SK Hynix's new US listing (SKHYV) surging 16% on debut, musing that retail enthusiasm may owe more to the ability to trade options than to the memory story itself. And he mapped the humanoid-robotics land-grab — pairings like Apptronik with Jabil (JBL) and Mercedes, plus a new Mitsubishi tie-up — noting how robotics firms keep coupling with major automakers, a demand vector he links to hyperscaler automation. CCXI appeared in that robotics list.

The throughline was consistent with his framework: conviction should track long-term revenue and operating-income growth, not the daily quote.